The visible map area is divided into a grid of cells. Each cell makes one Strava API call that returns the top 10 most popular segments in that area.
Higher density = smaller cells = finds more obscure segments that only appear when you zoom in close. But uses more API calls (limit: 200 per 15 minutes).
Cells that have already been scanned are cached and skipped for free. So scanning the same area twice costs nothing extra.